Overview
- He cautioned the future with AI is not guaranteed to be positive, calling powerful systems potentially destructive.
- He argued that forcing models to accept falsehoods can destabilize their reasoning, saying such pressure can make an AI “go insane.”
- The reliability problem was highlighted with recent iPhone alerts that fabricated a BBC sports result before Apple moved to clarify when Apple Intelligence generated notification text.
- The comments were made in a conversation with Zerodha co-founder Nikhil Kamath, placing Musk’s views within an ongoing public debate over AI safety.
- Context includes his departure from OpenAI’s board in 2018 and xAI’s 2023 launch of the Grok chatbot, as experts such as Geoffrey Hinton warn of a nontrivial chance of catastrophic outcomes.